NVIDIA Unveils NVLink Fusion System, Opens AI Chip Ecosystem

Ticker BuzzMonday, May 19, 2025 6:09 pm ET
2min read

NVIDIA, a prominent technology company, has unveiled a series of new technologies aimed at preserving its leading position in the development of artificial intelligence (AI). The company's CEO, Jensen Huang, announced these advancements during his keynote speech at a technology forum. The new technologies include the NVLink Fusion system, which allows customers to integrate their own central processing units (CPUs) with NVIDIA's AI chips or combine NVIDIA's CPUs with other manufacturers' AI accelerators. This move represents a significant departure from NVIDIA's previous strategy of providing systems based solely on its own components.

The NVLink Fusion system is designed to offer greater flexibility to data center customers while ensuring that NVIDIA remains at the forefront of technological innovation. This new approach allows for high-speed connections between processors and accelerators, providing a more customized and efficient AI infrastructure. Huang emphasized that this technology enables customers to tailor their AI infrastructure while keeping NVIDIA at the core.

In addition to the NVLink Fusion system, NVIDIA has also introduced several new AI chips, including the Grace Blackwell supercomputer and the Blackwell Ultra, which is set to be released later this year. The company also plans to launch the Feynman processor in 2028, continuing the legacy of its Rubin chip. These advancements are part of NVIDIA's broader strategy to transform from a chip manufacturer to a leader in AI infrastructure.

Huang highlighted that the computing industry, including NVIDIA itself, is undergoing a fundamental shift from a device-centric model to an AI factory-centric model. Data centers are becoming the computational units in this new paradigm. This transformation is driven by the increasing demand for AI computing power, which NVIDIA aims to meet with its comprehensive data center computing solutions.

The company's new technologies are not only aimed at maintaining its market leadership but also at fostering a more open and collaborative ecosystem. NVIDIA has partnered with several technology companies, including MediaTek, Marvell Technology Inc., and Alchip Technologies Ltd., to develop custom AI chips based on NVIDIA's processors. Additionally, Qualcomm and Fujitsu Limited plan to develop custom processors compatible with NVIDIA's computing system accelerators.

NVIDIA's strategy to open its AI chip ecosystem is a response to the growing trend of major clients, such as Microsoft and Amazon, designing their own processors and accelerators. By allowing customers to integrate their own components with NVIDIA's technology, the company aims to retain its dominance in the data center market while providing greater flexibility to its clients.

Huang's keynote speech also touched on the company's efforts to build a new headquarters, further expanding its technological production capabilities. He reflected on NVIDIA's journey from a graphics chip manufacturer to a leader in AI technology, driven by the rapid advancements in AI and the increasing demand for AI computing power.

In summary, NVIDIA's announcement of new technologies, including the NVLink Fusion system and several new AI chips, is a strategic move to maintain its leadership in the AI development and computing sectors. The company's shift towards an open and collaborative ecosystem, along with its comprehensive data center solutions, positions it well to meet the growing demand for AI computing power and retain its market dominance. The NVLink Fusion system, in particular, allows for greater flexibility and customization in AI infrastructure, ensuring that NVIDIA remains a key player in the AI landscape. This move is expected to strengthen NVIDIA's position in the market, as it continues to innovate and adapt to the evolving needs of its customers.